"- It simplifies hardware comparisons to a single value, which is its biggest flaw: it removes the aspect of the use case, which is very, very important in terms of comparing hardware. Sure, product A is 30% better than product B. At what? Making sandwiches?
- It includes a bunch of unweighted factors that are each of differing importance, and some of them are completely irrelevant. (remember when they used to account for things like market share and value? stupid)
- Testing methods are undisclosed, and so it's unclear how these results are actually taken.
- Impossible to actually validate benchmark results from a huge group of users. While it's nice to have such a large sample size, I'd rather have maybe twenty reputable reviewers with transparent testing methods than twenty thousand unknown samples that are unreliable."
"The problem with UBM is that it crowd-sources benchmark information from a user-base that doesn't always know what they're doing. Like if you look at RAM results, you might see 40% or more of people with XMP off judging by the score distribution. That makes it extremely inaccurate. As for 3D/GPU tests, it uses old DX9 and still suffers from the mob scoring due to random background processes that often affect the scores. For CPU, it literally calculates a certain % of background CPU usage and factors that into your results. The score changes literally every run."
"There is a variation of testing parameters. A user could be running completely thermal throttled, without a cooler at all, LN2 cooled, or undervolted. Ancient drivers that are no longer up to the task of what the system is asking to do.
All of these results are shared in the results pool and so a perfectly stable, fast 7700k system could potentially look terrible up against a delidded equivalent system with insanely tight RAM timings and 10c ambient."
  • It is usually safe to buy used CPUs, motherboards and GPUs
  • Used RAM is safe but is typically not cheaper than new prices
  • It is generally not advised to buy used storage drives and fans
  • Avoid used PSUs. They are an important part in your system that can kill other parts if the PSU is damaged/of poor quality.
  • eBay has a great Buyer Protection scheme that protects buyers from getting scammed when purchasing used parts.
An i5 9600K costs the same, if not more, than the Ryzen 5 3600.
The 9600K has 6 cores, 6 threads whilst the 3600 has 6 cores, 12 threads, therefore the 3600 has a noticeable increase in multi-core performance[1].
For games, the lack of threads could affect the 1% lows and have poor frametimes where the 3600 doesn't suffer from these issues. [2]
The 9600K does outperform the 3600 in most games by a very small amount [3], but this does not justify the loss of 6 threads. [4] At 1440p, the gap is almost non-existent. [5]
For anyone who isn't aiming for the highest possible overclock within safe temperatures, it really doesn't matter.
There is only a few degrees difference between most generic thermal pastes like Arctic MX-4 or Silver 5 and "high-end" thermal pastes like Thermal Grizzly Kryonaut for air coolers.
The difference is even less for liquid coolers.
However, when pushing a CPU to its limits, many factors influence the result so ensuring the best paste is used is optimal in such a case.
For most consumers, the brand of the RAM does not matter.
RAM modules are made by 3 companies: Samsung, Hynix and Micron.
Brands sometimes manufacture their own PCB, add their own heatsink, then sell it off. The difference between heatsinks are mostly just aesthetics. The quality of the PCB, for the most part, isn't relevant to anyone not interested in overclocking manually.
For the overclocking enthusiasts, some brands select specific die such as Samsung B-Die or Micron Rev E, that are great for overclocking further than XMP, for specific RAM kits. This should not matter to most people unless they are willing to spend time overclocking RAM.
For most users, 3200MHz C16 or 3600MHz C18 is the best choice.
Whilst performance gains beyond 3200MHz definitely exist in gaming and potentially in general use, they are usually not worth the increase in price of higher frequency kits.
Although CAS does matter, you want the lowest latency within a reasonable price. RAM that has a lower CAS latency than more typical models of the same bandwidth are typically far more expensive and the performance value drops significantly.
If you are into overclocking, you can currently purchase cheap 3000MHz C15/3200MHz C16 Micron Rev E, or 3600MHz 16-19-19 DJR RAM, that can overclock very well. If your budget is high you can even get cheap, good, Samsung B-Die at some stores in US and Europe/UK.
2 sticks of RAM is the optimal configuration in most systems, however if you know your motherboard has a T-Topology memory layout, 4 sticks will be better for overclocking.
If you are not concerned with overclocking past XMP and your XMP profile is within the specifications of 2 sticks under the QVL of your motherboard, 2 sticks would be preferred.
1st and 2nd generation Ryzen generally do not have strong enough IMCs to handle 4 sticks of high frequency RAM, therefore frequencies drop when using 4 sticks.
Using 2 sticks on 4 DIMM slot motherboards allows for upgrading in future if more RAM is needed.
Avoid single stick RAM as performance drops significantly (bandwidth halves).
Samsung SSDs are usually overpriced by a large amount.
Typical Samsung TLC SATA SSDs have little to no performance gains over other cheaper TLC SATA drives like MX500.
Samsung NVMe drives cost a lot more for a small performance gain, as seen in the graphs below. In some cases, lower capacity Samsung drives cost more than cheaper larger capacity drives.
There are cases where purchasing a Samsung drive is still sensible, but they are not relevant to most users especially on a small budget.
You should never consider an A320 motherboard, especially given how low B450 motherboard prices can get.
Quoted from Aegis#8622:
  • No VRM heatsinks
  • Low end audio codec
  • No debug features
  • Terrible BIOS support
  • No overclocking
  • Weak DIMM support
  • Lack of LLC function
  • Lack of ports
vgyoda#8008 and Homeslice#7282:
"Corsair/Logitech/Razer/HyperX and almost all other gamer boards suffer from poor build quality. Their thin plastic plates make keystrokes feel cheap and sound super pingy.
The same goes for the casing. The few plates and cases that are “aluminum” are thin coated airbrushed aluminum over thin plastics that still generally feel awful and scratch easily. The stabilizers, the things under your enter/backspace/space/shift keys, are awful in an objective manner. They’re built really really poorly and make the keyboard sound rattly and make those specific keys really wobbly. Not even the normal fixes that people do to improve their stabs (lubing, clipping, bandaid mod, padding, etc.) save 99% of gamer keyboards.
The keycaps are almost always in a proprietary layout and made with low quality ABS plastic. The legends are usually laser ablated, which means they ruin, wear and smudge fast and are almost impossible to replace due to the fact that their layout (or, in Logitech’s case, their stem) is proprietary.
Overall, keyboards branded and marketed towards gamers are inferior to lesser known products, which have superior build quality and customization overall."
"Response times are not a metric that consumers should worry about when selecting a monitor.
Many people confuse a monitor's response time with input lag, but response time is actually how quickly a pixel shifts from one color to another.
Although that still sounds like it would be important, there is no industry regulation on how to test response times, so what companies market are from the test that gave them the best results. Typically, companies market the time it takes for pixels to change from gray to gray, so the advertised numbers are almost always misleading. Although those values shouldn’t be used, it is true that too slow of a response time leads to ghosting/artifacting/pixel lag.
Consumers only have to worry about it in VA monitors where it is important to look for a panel with effective overdrive and black frame insertion which are features that counteract ghosting. MSI monitors and certain AOC ones are known to have respectable anti-ghosting tech."